Research map / Point clouds / Segmentation and detection
Active Perception using Light Curtains for Autonomous Driving
Siddharth Ancha, Yaadhav Raaj, Peiyun Hu, Srinivasa G. Narasimhan, David Held
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add first light curtain placed on uncertain regions | detector with single-beam LiDAR · Virtual KITTI | 3D mAP @0.5 IoU: 39.91 -> 58.01 | Improved |
| add first light curtain placed on uncertain regions | detector with single-beam LiDAR · SYNTHIA | 3D mAP @0.5 IoU: 60.49 -> 68.79 | Improved |
| add second and third light curtains | detector · Virtual KITTI | 3D mAP @0.5 IoU: 58.01 -> 60.86 -> 68.52 | Improved |
| add second and third light curtains | detector · SYNTHIA | 3D mAP @0.7 IoU: 55.99 -> 57.08 -> 57.30 | Improved |
| train jointly with light curtain data instead of single-beam only, evaluated on single beam (ablation) | detector · Virtual KITTI | 3D mAP @0.5 IoU: 42.35 -> 39.91 | Hurt |
| place curtains by dynamic programming instead of random depth | light curtain placement · Virtual KITTI | 3D mAP @0.5 IoU: 41.29 -> 58.01 | Improved |
| place curtains by dynamic programming instead of greedy optimization (ablation) | light curtain placement · Virtual KITTI | 3D mAP @0.5 IoU: 37.40 -> 58.01 | Improved |
| allow flexible curtain shapes instead of frontoparallel at max uncertainty depth (ablation) | light curtain placement · SYNTHIA | 3D mAP @0.5 IoU: 60.36 -> 68.79 | Improved |
| use uncertainty map instead of fixed depth 15m (ablation) | light curtain placement · Virtual KITTI | 3D mAP @0.5 IoU: 44.99 -> 58.01 | Improved |
| add online light curtain data generation during training (ablation) | detector training · Virtual KITTI | 3D mAP @0.5 IoU with 3 curtains: 0.68 -> 68.52 | Improved |
| add 10% noise to light curtain input | detector · Virtual KITTI | 3D mAP @0.7 IoU with 1 curtain: 35.29 -> 25.99 | Hurt |
| add 10% noise to light curtain input | detector · Virtual KITTI | 3D mAP @0.5 IoU with 3 curtains: 68.52 -> 60.02 | No clear effect |
Closest papers by meaning
- Agile Depth Sensing Using Triangulation Light Curtains (ICCV 2019 · similarity 0.87)
- Programmable Triangulation Light Curtains (ECCV 2018 · similarity 0.86)
- Exploiting & Refining Depth Distributions With Triangulation Light Curtains (CVPR 2021 · similarity 0.86)
- Robust 3D Object Detection using Probabilistic Point Clouds from Single-Photon LiDARs (ICCV 2025 · similarity 0.86)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.84)
- STONE: A Submodular Optimization Framework for Active 3D Object Detection (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
